How we judge it

Start with the job to be done, then record success criteria, time, human corrections, and failure boundaries. Feature count is never a proxy for outcome quality.

  • The same input and constraints
  • A reproducible workflow
  • Explicit failure examples

Turn answers into evidence

Map every important claim to its original source, date, subject, and scope. A citation that cannot lead back to primary material is a lead, not evidence.

Decision

A useful verdict names who it fits, when not to use it, and the cost and data boundaries to verify before adoption.